Hacker News Reader

Comment by bschne | Hacker News Reader

(how) do you (try to) correct for this with interviews?

Someone's bound to mention it in this thread (or I hope they do, because I thought it was great but can't remember the name or the details) - but there's some formula that's roughly like:

  N = total applicants
  1. Interview N/10, decline no matter what 
  2. Hire the next person who's better than everyone seen so far

As I recall it was slightly more complex than that, perhaps only to give it an impression of being rigorous theory, but that was the gist of it.

I think you're looking for the Secretary problem. And the solution is almost what you said, but instead of N/10 it's a bit over N/3.

This only applies if you have to make a hire/skip decision immediately after each interview and can't go back later. (The formulation I usually see is a princess choosing a suitor — for pride/face reasons, she can't pass on a suitor and then go back if it turns out he was the best of the bunch.) That's not typically how job interviews work, though.

I think the magic number is 1/e of the applicant pool (e = Napier's constant, 2.7...)

That mathematical problem (the secretary problem) is often cited, but it's not how real hiring works. The problem defines the only success criterion as hiring the single top candidate, and everything else fails. That's not at all what you're looking for in the real world - there are a spectrum of results, where any one in the top decile will be great and the next decile only slightly lesser and so on.

One-dimensional ranking is one aspect of how it breaks in real-world hiring, but I think almost the bigger might be that the secretary problem assumes you need to decide immediately and can't just wait around until you've seen the whole batch of applicants.

The main difference is that you're trying to solve the problem of turning an interview into a way to choose between candidates, while the secretary problem assumes you already figured that out and is trying to solve something different entirely!

This is the secretary problem [1] from optimal stopping theory, but it doesn't apply to hiring because you can wait until the end to make a decision once you've seen everyone.

The main problem here indicated by OP isn't explore/exploit, it's that the ordering of candidates seemingly distorts their rating of them.

1. https://en.wikipedia.org/wiki/Secretary_problem

Aka 37% rule

Reply on news.ycombinator.com